In constrained least squares based super-resolution image reconstruction, it is an important issue to up-sample and merge the high-frequency information contained in low resolution images efficiently and without artifact. In this paper, we analyze up-sampling methods used for registering low resolution data to high resolution grid and show their advantages and disadvantages. In addition, we propose a new up-sampling and merging method called "Hybrid up-sampling". The proposed method accurately incorporates the high frequency data in low resolution images and minimizes the artifact caused by data deficiency. By this method, regions that the registered high frequency data do not cover are naturally regularized without using complex regularizes. Experimental results show that the choice of up-sampler significantly affects the quality of resulting images and the proposed up-sampler gives better results compared to conventional simple up-samplers, especially when the number of low resolution images is deficient.